Meta introduces Muse Image and Muse Video

Background

Muse Image and Muse Video were the first media generation models from Meta Superintelligence Labs. They followed Muse Spark, the lab's first language and reasoning model.^[1] Media generation models create pictures, video or audio, while a language and reasoning model handles text and multi-step problems. The two kinds of model can work together, as described below.

What happened

On 7 July 2026 Meta Superintelligence Labs introduced Muse Image and Muse Video, its first media generation models.^[1] Muse Image became available in the Meta AI app and on meta.ai. It also reached Instagram Stories in the United States and WhatsApp in some countries.^[1]

Key details

Muse Video. Muse Video is built on the same base as Muse Image and adds native audio. It entered early preview.^[1] Native audio means sound is produced by the same model as the picture, rather than being added by a separate tool.

Rankings. Meta said Muse Image ranked No. 2 on Arena for text-to-image and image editing, and Muse Video ranked No. 3 for text-to-video. Meta gave these positions as of 5 July 2026.^[1] Arena is a public leaderboard where people compare outputs from different models, so the ranking reflects human votes rather than a fixed test.

Watermark. Images made with Muse Image carry an invisible "Content Seal" watermark.^[1] An invisible watermark is built into the image itself and is not meant to be seen by a viewer.

Working with Muse Spark. Muse Image works with Muse Spark and can use tools such as search and code.^[1] This means a user can ask the language model to plan or search for something, and the image model can then use that work to make the picture.

Impact

The release extended the Muse series from text and reasoning into images and video. Because Muse Image reached Instagram Stories and WhatsApp as well as Meta's AI app, its image generation was placed inside the apps that people already use. Images made with the model carry the invisible "Content Seal" watermark.^[1]